TreemapChart

Root hierarchy node. Children are nested arbitrarily deep.

Container height in pixels. Width is always responsive.

Item selection. true = single selection keyed by the default key.

Accessible name, description, data table and keyboard navigation. No announcements: it switches <Chart>'s interaction announcements, which this family does not make.

The entrance: on, off, duration, easing and the family's entrance type (ADR-082, LAB-AT17).

Optional className applied to the outer container.

Colour the tiles by a measure (T5): each leaf's own colorValue through one scale over the whole tree, with a colour key, while value keeps sizing it. Opt-in; without it every branch keeps its colour. The group headers keep their branch colours; a tile without a colour value is drawn in a neutral "no value" paint. See ColorMeasureOptions.

Categorical palette for this chart only: replaces the host palette (--chart-1 … --chart-10) in the chart's colour assignment, and nothing else. It is a data encoding, not a theme: surfaces, text, grid and semantic roles keep coming from the CSS token contract, and a series, datum or annotation that sets its own colour keeps it. Any CSS colour string (a public token reference such as var(--chart-3), hex, oklch).

The context menu. Opened on a node, the host's builder receives a typed HierarchyElement (node, depth, path; the deepest node under the pointer) as context.target.element.

The initial selection of an uncontrolled chart.

The chart's UI direction: the reading direction of its HTML text and chrome (tooltip, legend, menus, the accessible table), and the direction the context menu carries under <body>.

Left out, the chart inherits the direction of the page around it (the nearest dir attribute, else the computed CSS direction) and follows a later change of that attribute. It is never inferred from locale, and it never reverses an axis or reorders data: reversed stays an explicit axis choice. Under rtl the chrome mirrors (legend flow, tooltip side, context menu and submenus, report text in exports), a title-like label starts at the inline start, and a label placed by geometry (an axis tick, a slice or bar label) keeps its physical side; plot arrow keys stay spatial while legend and menu arrows follow the reading direction.

Enable click-to-zoom drilldown into child nodes. Default: false.

The chart's events, and the selection events (T8): an item is a LEAF, named by its id ({kind: 'node', id}); a group is structure and keeps its header's drilldown. A leaf without an id is activated but never selected.

The label policy (T6): a tile's second line (secondary, default 'share': its share of the shown total; 'value': its value in valueSemantics; 'none') and the levels whose group headers and tiles draw text (levels, counted from the shown root). Visual only: the tooltip, the table and the announcements keep every name and value. See HierarchyLabelOptions.

Presentation input for axis labels, tooltips, labels and accessibility announcements: a locale code (notation follows the language), or a presentation-context object ({languageLocale, dateShape?, dateFormat?, timeFormat?, numberFormat?, currencyDisplay?}) carrying language and notation independently. dateShape is the normal human date shape a plain date consumer inherits; an axis that declares its own preset for readability keeps it. Forwarded untouched to the central @workspace/active-format; the package never resolves tenant or application settings itself. When omitted, <ChartDefaults> supplies it; with neither, the chart throws a configuration error (no 'en' fallback, active-format D14). Object inputs must be referentially stable across renders.

Padding between cells in pixels. Default: 2.

Padding around parent groups in pixels. Default: 1.

Imperative handle: exports and element access. A family built on this type has no data-range zoom, so its handle has no setZoom. Allows getting a ref to the component instance. Once the component unmounts, React will set ref.current to null (or call the ref with null if you passed a callback ref).

The selection, controlled by the host. The chart draws, announces and lists the items it resolves to and never changes it; the host updates it from onSelectionChange. A request the host applies is announced when the host passes it back; one it ignores is never announced. A reference to an item that is hidden, zoomed or drilled out of view, or not in the data stays in the host's selection and is not shown as selected.

The host translator for every text the chart shows or announces. When omitted, <ChartDefaults> supplies it; with neither, the chart throws a configuration error (no English fallback, ADR-028).

Tooltip of the node under the pointer: enabled, renderer, valueFormatter and style (ADR-082, LAB-AT17).

What the node values mean (BC1): unit and precision, declared once with the type the Cartesian charts read (ChartValueSemantics). The tooltip, the data table and the announcement write a node's value in it; tooltip.valueFormatter still wins in the tooltip. A tile shows its share of the shown total, which stays a percentage. Without a declaration the chart presentation's number applies. CSV and Copy value keep the raw number.

Optional fixed width in pixels. Default: fill container.

Fired when the reader changes the selection (pointer, touch or keyboard), and when an uncontrolled selection drops items the data no longer contains. selection is the complete selection the change makes; it keeps every reference the selection held, of whatever form. On a controlled chart it is a proposal (change.requested: true), computed from the current selectedItems and never from an earlier proposal; the host applies it, changes it or ignores it.

Fired when a chart animation completes.

• 'load': initial mount animation finished (or immediately if animation disabled / prefers-reduced-motion is active).
• 'update': data-update transition finished.

Not triggered by zoom animations (use onZoomChange) or tooltip fades.

Fired after a successful export action (PNG/SVG/CSV/PDF), including from the context menu and from imperative ChartHandle calls.

Fired when an export could not produce its file (W11), from the handle as well as from the context menu. Without it a failed context-menu export is reported as an error diagnostic (<ChartDefaults diagnostics>).

Fired when the reader activates an item: a click or tap on it, or Enter on the focused item (Shift+Enter on a sankey, whose Enter moves between nodes and links). With a handler, Enter activates instead of opening the context menu (Shift+F10 and the context-menu key still open it).

Fired exactly once per chart instance, after the first successful mount with valid dimensions and before any entrance animation starts. "Ready" means: the SVG is in the DOM, layout has run with real dimensions, and the chart can be queried by integration tests.

Distinct from onAnimationComplete with type: 'load', which fires *after* the load animation finishes. A chart with animation={false} still fires onReady — animation state has no effect on readiness.

A new React key on the host produces a new component instance and onReady fires again. Identity changes of the callback itself after the first fire are ignored (latch is per-instance).

Consumer requirement — stable from first render. The latch closes on the first effect run regardless of whether onReady was defined at that moment. If onReady is initially undefined and later resolves to a function, the callback will never fire — the latch has already closed with a no-op. Pass a stable reference (typically via useCallback) or pass the final function unconditionally from the first render.